Question 2 Background High blood pressure, or hypertension, occurs when a person's blood pressure increases to unhealthy levels. Blood pressure measurement takes into account how much blood is passing through the blood vessels and the amount of resistance the blood meets while the heart is pumping. High blood pressure can damage a person's arteries by making them less elastic, which decreases the flow of blood and oxygen to the heart and leads to heart disease (refer to https://www.cdc.gov/bloodpressure/about.htm). MyCare is one of the main medical institutions in Singapore providing treatment for people suffering from high blood pressure. Like most healthcare providers, one of the primary objectives of MyCare is to improve the quality of life of patients and reduce healthcare costs through more effective diagnoses and treatments. Hence, MyCare has embarked on the use of data mining to profile patients with mild high blood pressure (hence not requiring long-term mediation) based on their demographics and lifestyle, and track whether their blood pressure has improved within 2 years. A dataset containing 1300 patient records (with 685 patients or 52.7% showing an improvement in their blood pressure measurement within 2 years) is extracted from the database for analysis. The description of the dataset given in Table 1. Table 1. Description of MyCare's Dataset Variable Name Variable Description Possible Values PatID Patient Identification Number 0001 to 1300 Gender Gender Female Male Age Age 30 to 70 WeightStatus Weight status of the patient based on his/her Underweight body mass index (BMI) before the patient Normal suffers from high blood pressure Overweight Exercise Number of times that the patient exercises in a 0 to 7 week Smoke Number of cigarettes that the patient smokes 0 to 150 in a week Drink Number of beverages with alcohol (500ml) that | 0 to 15 the patient drinks in a week Caffeine Number of beverages with caffeine (200ml) 0 to 15 that the patient drinks in a weekFamilyHist Whether the patient has any immediate family Yes members who also have high blood pressure. No Improve Whether the patient's blood pressure Yes measurement has improved within 2 years. No TwoStep clustering was used to mine the data. The clustering solution is given in Appendix B. Based on the information given and the solution provided in Appendix B, answer the following questions: (a) Assess the quality clustering solution and interpret the cluster profiles. (17 marks) (b) Based on the first three most important attributes of the clustering results, propose the descriptive labels for the three clusters. (3 marks) (c) Based on the cluster profiles discussed in (a), appraise the deployment of the findings by providing two recommendations on how the clustering solution can be used to improve the blood pressure of the patients. (10 marks) (d) MyCare would also like to use CHAID, a decision tree algorithm, to predict who is likely to show an improvement in blood pressure measurement. Explain in less than 300 words how you would implement the data mining project with respect to the phases of the CRISP-DM framework. (Note: Any words beyond 300 words will not be marked.) (15 marks) (Note: The above findings are for assessment purposes only and do not comprise medical advice or actual treatments.)Appendix B Table 1: Settings of Variables in SPSS Modeler Data Filter Types Annotations Read Values Clear Values Clear All Values Field Measurement Values Missing Check Role PalD Continuous [1.0,1300.0] None id Record ID Gender Flag Male/Female None Input A Age Continuous [30.0,68.0] None Input A WeightStatus Nominal Normal, Overweight,... None Input Exercise Continuous [0.0,7.0] None Input Smoke Continuous [0.0, 140.0] None Input Drink Continuous [0.0,15.0) None Input Caffeine Continuous [0.0,15.0] None Input A FamilyHist Flag Yes/No None Input Improve Flag Yes/No None NoneCluster Sizes Cluster cluster-1 cluster-2 Model Summary cluster-3 21.5% Algorithm TwoStep 48.6% Inputs B 29.9% Clusters 3 Cluster Quality Average Silhouette= 0.5 Poor Fair Good 05 D.O D.5 10 Size of Smallest Cluster 279 (21.5%) Silhouette measure of cohesion and separation Size of Largest Cluster 632 (48.6%) Ratio of Sizes: Largest Cluster to 2.27 Smallest Cluster Figure 1. Model SummaryClusters Input (Predictor) Importance 1.0 0.8 0.6 0.4 0.20.0 Cluster cluster-1 cluster-2 cluster-] Label Description Size 18.6% 29.9%% 21.5% (632) (389) (279) Inputs Age Age Age Exercise Exercise Exercise Smoke Smoke Smoke WeightStatus WeightStatus WeightStatus Gender Gender Gender FamilyHist FamilyHist FamilyHist Caffeine Caffe ine Caffeine Drink Drink Drink Figure 2. Cluster Relative DistributionCluster Comparison cluster-1 cluster-2 _ cluster-3 Age Exercise Smoke Weight Status Normal Overweight Underweight Gender Female Male FamilyHist No Yes Caffeine Drink Figure 3. Cluster Comparison Table 3. Matrix of Cluster ID and ImproveImprove ST-TwoStep No Yes Total cluster-1 Count 395 237 632 Row % 62 500 37.500 100 Column % 64.228 34.599 48.615 Total % 30.385 18.231 48.615 cluster-2 Count 203 186 389 Row % 52 185 47.815 100 Column % 33.008 27.153 20,923 Total % 15.615 14.308 20.923 cluster-3 Count 17 262 279 Row % 6.093 93.907 100 Column % 2.764 38.248 21.462 Total % 1.308 20.154 21.462 Total Count 615 685 1300 Row % 47.308 52.692 100 Column % 100 100 100 Total % 47.308 52.692 100